analyze(_:atAudioFramePosition:)
Adds a new audio buffer to the analyzer’s larger stream buffer.
Declaration
func analyze(_ audioBuffer: AVAudioBuffer, atAudioFramePosition audioFramePosition: AVAudioFramePosition)Parameters
- audioBuffer:
The audio data buffer.
- audioFramePosition:
The frame position of the data in the buffer. The analyzer expects this parameter to monotonically increase with each call. Otherwise, the analyzer may reset its internal state to account for the jump in time.
Discussion
Serialize all calls to this method on a dedicated dispatch queue to prevent blocking the calling thread. The audio stream analyzer sends errors to each request’s results observer.
The method handles audio buffers that vary in size. If necessary, the analyzer regroups the data to a block size the underlying Core ML model expects. The analyzer uses a fixed-size audio block from some audio analysis types and may call a request’s results observer one time or many times. The factors that affect the number of calls are:
The input buffer size
The underlying model’s native analysis block size
The analyzer’s current state
By default, the analyzer processes data on the first audio channel in the audio stream. The analyzer converts the sample rate of the data if it doesn’t match the underlying model’s requirements.